Synology DS925+, two failed drives: 99% recovered
3 × 4 To · 2026 · Mons
Failure : A company near Mons kept all its project folders on a Synology DS925+ NAS with no backup at all: four 4 TB Synology HAT3300-4T drives in SHR-1. A first drive failed, and its rebuild stopped at 0.4%. Shortly after, a second drive was ejected and the volume went offline. Synology support tried reinstalling DSM, failed, and closed the ticket. The company's IT partner then brought us the three remaining drives. One of them reported SMART FAILED, with 13,040 pending sectors, and was reading at under 1 MB/s.
Work done : Every drive was imaged first; all later work ran on the copies. Analysis showed two four-member mdadm RAID 5 arrays stacked under LVM. Assembly failed at first: the drive order had been changed during earlier attempts, and the exact position of each member had to be worked out. On the failing drive, head 3 (of 4) was reading very poorly. Rather than wear it down until it died, we opened it in the cleanroom, replaced the head stack with a donor set from our own stock, and imaged the drive with PC-3000. The 2026 folder was the most urgent, so it came out first.
Result : The volume could be rebuilt because the most damaged drive was read with healthy heads while it still worked. 99% of the 1.6 TB of files came back; the only losses were 304 empty files (0 bytes), all listed in the report handed to the client.
Synology DS223, RAID 1 wiped by DSM: 100% recovered
4 To · 2026 · Bruxelles
Failure : A computer repair shop in Brussels took in a customer's NAS: a Synology DS223 with two 4 TB WD Red drives in a mirror. The unit had been warning for days that one drive needed replacing. The technician swapped the failing drive and, on restart, DSM offered to install itself fresh. He accepted. The installation reinitialised the volume: the data partition was gone and the mirror had nothing left to rebuild. The shop owner called us the same day. We asked him to bring both drives, including the old one that most people throw away after a swap.
Work done : We worked on the drive the NAS had dropped, and never plugged the enclosure back in: powering the DS223 up again would have let DSM write over the data. The old WD40AFAX was not dead. It had unreadable sectors clustered in one area, enough for Synology to call it failed, not enough to stop it being read. We imaged it sector by sector on PC-3000 Data Extractor, skipping the unstable areas first, then going back over them with longer timeouts until the copy was complete. The mirror was then rebuilt from that single image and the Synology file system mounted to pull the folders out. In RAID 1 both drives hold the same data, which is why the invoice covers one cloned drive, not two.
Result : The volume had been reinitialised, not physically erased: the data was still on the platters and the rejected drive was still readable. We returned 100% of the end customer's files on an external drive, four days after drop-off.
Synology BeeStation dropped: 300 GB of photos saved
4 To · 2026 · Bruxelles
Failure : Every photo a Brussels family had lived in one place: a 4 TB Synology BeeStation BST150 standing upright on a shelf. There was no second copy anywhere, because the station was itself the backup for the phones and the home computer. During building work someone knocked the shelf and the narrow, vertical unit fell on its side. After that the BeeStation would not start: no network access, nothing normal at power-up. The owner brought it to us without having tried to switch it on again and again, and that counted for a lot later.
Work done : Under the Synology label sat a 4 TB Seagate drive with media cache, taken out of the enclosure in the lab. The fall had damaged the heads: the drive could no longer read its service area and never reached ready state. Firmware access was locked too, and that is where most of the difficulty was.
We unlocked the ROM on the board, rewrote it and adjusted the drive's passport settings so it would start in technological mode without touching the damaged areas. A head map was then built: the failed heads excluded from reading, the good ones doing the work. The point was not to wear them out for nothing. Instead of a full 4 TB image, we located the partition, analysed its structure and read only the areas holding the photos, 300 GB in all. The reading itself ran over a week, at reduced speed and in several passes, to spare heads that were already weakened.
Result : The fall hit the heads, not the surface where the data lived, and that is what makes the result complete. All 300 GB of family photos were returned at 100%, with no missing files, on a new drive handed to the client.